|
CIM based Cluster Management
The intention of the cooperation between the University of Illinois/NCSA and the SBLIM Project is to show the value of Linux based GRID computing for academic and business communities and to advance these technologies. The goal of the SBLIM development project with NCSA's High Performance Computing and Communications Systems Group is the implementation of a CIM based concept for monitoring a High Performance Computing (HPC) Linux Cluster. Both SBLIM & NCSA will benefit from this cooperative effort as we are advancing GRID Computing and CIM technologies. Both organizations bring a set of critical skills. The NCSA offers us the HPC skill and environment needed to develop, implement and test a CIM based HPC monitoring concept. SBLIM will offer the NCSA a scalable HPC monitoring solution consisting of
CIM based Distributed Monitoring In the GMA versus CIM paper Adrian Schuur compares the GRID Monitoring Architecture with possible CIM models for monitoring a Cluster. Adrian Schuur and Viktor Mihajlovski also collected some thoughts about CIM based Distributed Monitoring in this second referenced document. Managing OpenPBS with CIM OpenPBS is an Open Source Product to submit and manage Jobs to / in a Cluster. Monitoring Jobs in a Cluster consists of two major components. The first one is getting a picture of the "whole", while the other component is the local management and monitoring of each single system. OpenPBS is responsible for monitoring the whole picture. The Management of OpenPBS with CIM document describes the latest version of the JSIM based OpenPBS CIM Model. The local management of the cluster nodes is done by using the SBLIM Base Instrumentation. A set of the available providers is described in the Functional Specification. The old version of the CIM based Cluster Management Proposal is still available. |
|||||||||||||||||||||||||||
|